Fundamental AI

NIST’s AI portfolio includes fundamental research, evaluations, and standards for AI technologies – including software, hardware, architectures, human- interaction and teaming, and all relevant intersections and interfaces – vital for AI computational trust.

Establishing and Promoting Sociotechnical Requirements for Trustworthy and Responsible AI

• Through targeted research investments, NIST advances standards, tests, and evaluations for  mapping, measuring, and managing development of trustworthy AI and its responsible use. These include developing  taxonomy, terminology, testbeds and benchmarks for measuring AI risks – as well as standards needed for key sociotechnical characteristics of AI trustworthiness. Trustworthy AI systems are demonstrated to be valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair with harmful bias managed.
• By convening stakeholder workshops and via other engagements, NIST is developing a resource center of documents, software, standards, and related tools that contribute to better understanding, identifying, measuring, and managing various risks associated with AI systems.
• Reflecting private sector interest and as mandated by Congress, NIST developed a voluntary AI risk management framework through collaboration with stakeholders across public and private sectors.

Hardware for AI: Creating New Measurements and Technical Approaches for New AI Chips

• Demand for faster, more energy-efficient information processing is growing exponentially as AI becomes more prevalent in our everyday lives. Conventional digital processing hardware cannot keep up with this demand. That is why researchers, taking inspiration from the brain, are considering alternatives where massively connected networks of artificial neurons and synapses process information with high speed, energy efficiency, scalability, and adaptive learning capabilities. 
• To that end, NIST is helping to develop devices, circuits, systems, measurements, and theory to support the evolution of AI hardware technology from laboratory research to commercial applications. It is focusing on scalability, energy efficiency, hardware optimization, and architecture development. The new hardware leverages the physics of devices to perform computations, while the architectures and algorithms have entirely new intelligent functionality. This calls for a new system of measurement techniques and protocols.
  For more information about this work, see Hardware for AI.